--experimental-test-coverage falsely reports missing coverage where TS source is import type #54753
Description
Version
v22.8.0
Platform
Darwin TRI-N93DLJDY6Y 23.6.0 Darwin Kernel Version 23.6.0: Mon Jul 29 21:14:30 PDT 2024; root:xnu-10063.141.2~1/RELEASE_ARM64_T6000 arm64
Subsystem
No response
What steps will reproduce the bug?
With this TypeScript source code in src/a.mts (where the import is from doesn't matter):
import type {} from "node:assert";
console.log("Hi");Compiling that to dist/a.mjs:
console.log("Hi");
export {};
//# sourceMappingURL=a.mjs.mapWith source map dist/a.mjs.map:
{"version":3,"file":"a.mjs","sourceRoot":"","sources":["../src/a.mts"],"names":[],"mappings":"AAEA,OAAO,CAAC,GAAG,CAAC,IAAI,CAAC,CAAC"}
Then in test.mjs:
import "./dist/a.mjs";Run:
node --experimental-test-coverage --test test.mjsHow often does it reproduce? Is there a required condition?
Every time.
What is the expected behavior? Why is that the expected behavior?
100% code coverage reported for the module src/a.mts.
What do you see instead?
Notice the false missing line coverage reported in the terminal output:
Hi
✔ test.mjs (49.142ms)
ℹ tests 1
ℹ suites 0
ℹ pass 1
ℹ fail 0
ℹ cancelled 0
ℹ skipped 0
ℹ todo 0
ℹ duration_ms 53.724042
ℹ start of coverage report
ℹ ----------------------------------------------------------
ℹ file | line % | branch % | funcs % | uncovered lines
ℹ ----------------------------------------------------------
ℹ src/a.mts | 66.67 | 100.00 | 100.00 | 1
ℹ test.mjs | 100.00 | 100.00 | 100.00 |
ℹ ----------------------------------------------------------
ℹ all files | 75.00 | 100.00 | 100.00 |
ℹ ----------------------------------------------------------
ℹ end of coverage report
Additional information
If you comment out the import type {} from "node:assert"; and rebuild, a second run for functionally the same runtime code now correctly reports no missing coverage:
Hi
✔ test.mjs (47.03ms)
ℹ tests 1
ℹ suites 0
ℹ pass 1
ℹ fail 0
ℹ cancelled 0
ℹ skipped 0
ℹ todo 0
ℹ duration_ms 51.619166
ℹ start of coverage report
ℹ ----------------------------------------------------------
ℹ file | line % | branch % | funcs % | uncovered lines
ℹ ----------------------------------------------------------
ℹ src/a.mts | 100.00 | 100.00 | 100.00 |
ℹ test.mjs | 100.00 | 100.00 | 100.00 |
ℹ ----------------------------------------------------------
ℹ all files | 100.00 | 100.00 | 100.00 |
ℹ ----------------------------------------------------------
ℹ end of coverage report
If you don't create source maps when compiling TypeScript modules containing import type, the Node.js test runner correctly reports no missing coverage. So the problem is around how Node.js is interpreting the source maps. Only runtime code should determine code coverage; not source code like TypeScript import type that is eliminated in the build.
Something else that is strange, is that the Node.js CLI flag --enable-source-maps doesn't seem to have an effect on how the Node.js test runner reports coverage; even without the flag it will always take into account the source maps information. Why is coverage exempt from respecting how --enable-source-maps works for other Node.js features?